tcg/s390x: Add host vector framework

Add registers and function stubs.  The functionality
is disabled via squashing s390_facilities[2] to 0.

We must still include results for the mandatory opcodes in
tcg_target_op_def, as all opcodes are checked during tcg init.

Reviewed-by: David Hildenbrand <[email protected]>
Signed-off-by: Richard Henderson <[email protected]>

Author: rth7680

tcg/s390x/tcg-target-con-set.h

@@ -13,13 +13,17 @@ C_O0_I1(r)
 C_O0_I2(L, L)
 C_O0_I2(r, r)
 C_O0_I2(r, ri)
+C_O0_I2(v, r)
 C_O1_I1(r, L)
 C_O1_I1(r, r)
+C_O1_I1(v, r)
+C_O1_I1(v, vr)
 C_O1_I2(r, 0, ri)
 C_O1_I2(r, 0, rI)
 C_O1_I2(r, 0, rJ)
 C_O1_I2(r, r, ri)
 C_O1_I2(r, rZ, r)
+C_O1_I2(v, v, v)
 C_O1_I4(r, r, ri, r, 0)
 C_O1_I4(r, r, ri, rI, 0)
 C_O2_I2(b, a, 0, r)

tcg/s390x/tcg-target-con-str.h

@@ -10,6 +10,7 @@
  */
 REGS('r', ALL_GENERAL_REGS)
 REGS('L', ALL_GENERAL_REGS & ~SOFTMMU_RESERVE_REGS)
+REGS('v', ALL_VECTOR_REGS)
 /*
  * A (single) even/odd pair for division.
  * TODO: Add something to the register allocator to allow

tcg/s390x/tcg-target.c.inc

@@ -43,6 +43,8 @@
 #define TCG_CT_CONST_ZERO  0x800
 
 #define ALL_GENERAL_REGS     MAKE_64BIT_MASK(0, 16)
+#define ALL_VECTOR_REGS      MAKE_64BIT_MASK(32, 32)
+
 /*
  * For softmmu, we need to avoid conflicts with the first 3
  * argument registers to perform the tlb lookup, and to call
@@ -268,8 +270,13 @@ typedef enum S390Opcode {
 
 #ifdef CONFIG_DEBUG_TCG
 static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = {
-    "%r0", "%r1", "%r2", "%r3", "%r4", "%r5", "%r6", "%r7",
-    "%r8", "%r9", "%r10" "%r11" "%r12" "%r13" "%r14" "%r15"
+    "%r0",  "%r1",  "%r2",  "%r3",  "%r4",  "%r5",  "%r6",  "%r7",
+    "%r8",  "%r9",  "%r10", "%r11", "%r12", "%r13", "%r14", "%r15",
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    "%v0",  "%v1",  "%v2",  "%v3",  "%v4",  "%v5",  "%v6",  "%v7",
+    "%v8",  "%v9",  "%v10", "%v11", "%v12", "%v13", "%v14", "%v15",
+    "%v16", "%v17", "%v18", "%v19", "%v20", "%v21", "%v22", "%v23",
+    "%v24", "%v25", "%v26", "%v27", "%v28", "%v29", "%v30", "%v31",
 };
 #endif
 
@@ -295,6 +302,32 @@ static const int tcg_target_reg_alloc_order[] = {
     TCG_REG_R4,
     TCG_REG_R3,
     TCG_REG_R2,
+
+    /* V8-V15 are call saved, and omitted. */
+    TCG_REG_V0,
+    TCG_REG_V1,
+    TCG_REG_V2,
+    TCG_REG_V3,
+    TCG_REG_V4,
+    TCG_REG_V5,
+    TCG_REG_V6,
+    TCG_REG_V7,
+    TCG_REG_V16,
+    TCG_REG_V17,
+    TCG_REG_V18,
+    TCG_REG_V19,
+    TCG_REG_V20,
+    TCG_REG_V21,
+    TCG_REG_V22,
+    TCG_REG_V23,
+    TCG_REG_V24,
+    TCG_REG_V25,
+    TCG_REG_V26,
+    TCG_REG_V27,
+    TCG_REG_V28,
+    TCG_REG_V29,
+    TCG_REG_V30,
+    TCG_REG_V31,
 };
 
 static const int tcg_target_call_iarg_regs[] = {
@@ -377,7 +410,7 @@ static void * const qemu_st_helpers[(MO_SIZE | MO_BSWAP) + 1] = {
 #endif
 
 static const tcg_insn_unit *tb_ret_addr;
-uint64_t s390_facilities[1];
+uint64_t s390_facilities[3];
 
 static bool patch_reloc(tcg_insn_unit *src_rw, int type,
                         intptr_t value, intptr_t addend)
@@ -2293,6 +2326,42 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc,
     }
 }
 
+static bool tcg_out_dup_vec(TCGContext *s, TCGType type, unsigned vece,
+                            TCGReg dst, TCGReg src)
+{
+    g_assert_not_reached();
+}
+
+static bool tcg_out_dupm_vec(TCGContext *s, TCGType type, unsigned vece,
+                             TCGReg dst, TCGReg base, intptr_t offset)
+{
+    g_assert_not_reached();
+}
+
+static void tcg_out_dupi_vec(TCGContext *s, TCGType type, unsigned vece,
+                             TCGReg dst, int64_t val)
+{
+    g_assert_not_reached();
+}
+
+static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc,
+                           unsigned vecl, unsigned vece,
+                           const TCGArg *args, const int *const_args)
+{
+    g_assert_not_reached();
+}
+
+int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece)
+{
+    return 0;
+}
+
+void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece,
+                       TCGArg a0, ...)
+{
+    g_assert_not_reached();
+}
+
 static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op)
 {
     switch (op) {
@@ -2433,11 +2502,34 @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op)
                 ? C_O2_I4(r, r, 0, 1, rA, r)
                 : C_O2_I4(r, r, 0, 1, r, r));
 
+    case INDEX_op_st_vec:
+        return C_O0_I2(v, r);
+    case INDEX_op_ld_vec:
+    case INDEX_op_dupm_vec:
+        return C_O1_I1(v, r);
+    case INDEX_op_dup_vec:
+        return C_O1_I1(v, vr);
+    case INDEX_op_add_vec:
+    case INDEX_op_sub_vec:
+    case INDEX_op_and_vec:
+    case INDEX_op_or_vec:
+    case INDEX_op_xor_vec:
+    case INDEX_op_cmp_vec:
+        return C_O1_I2(v, v, v);
+
     default:
         g_assert_not_reached();
     }
 }
 
+/*
+ * Mainline glibc added HWCAP_S390_VX before it was kernel abi.
+ * Some distros have fixed this up locally, others have not.
+ */
+#ifndef HWCAP_S390_VXRS
+#define HWCAP_S390_VXRS 2048
+#endif
+
 static void query_s390_facilities(void)
 {
     unsigned long hwcap = qemu_getauxval(AT_HWCAP);
@@ -2452,6 +2544,16 @@ static void query_s390_facilities(void)
         asm volatile(".word 0xb2b0,0x1000"
                      : "=r"(r0) : "r"(r0), "r"(r1) : "memory", "cc");
     }
+
+    /*
+     * Use of vector registers requires os support beyond the facility bit.
+     * If the kernel does not advertise support, disable the facility bits.
+     * There is nothing else we currently care about in the 3rd word, so
+     * disable VECTOR with one store.
+     */
+    if (1 || !(hwcap & HWCAP_S390_VXRS)) {
+        s390_facilities[2] = 0;
+    }
 }
 
 static void tcg_target_init(TCGContext *s)
@@ -2460,6 +2562,10 @@ static void tcg_target_init(TCGContext *s)
 
     tcg_target_available_regs[TCG_TYPE_I32] = 0xffff;
     tcg_target_available_regs[TCG_TYPE_I64] = 0xffff;
+    if (HAVE_FACILITY(VECTOR)) {
+        tcg_target_available_regs[TCG_TYPE_V64] = 0xffffffff00000000ull;
+        tcg_target_available_regs[TCG_TYPE_V128] = 0xffffffff00000000ull;
+    }
 
     tcg_target_call_clobber_regs = 0;
     tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R0);
@@ -2474,6 +2580,31 @@ static void tcg_target_init(TCGContext *s)
     /* The return register can be considered call-clobbered.  */
     tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_R14);
 
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V0);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V1);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V2);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V3);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V4);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V5);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V6);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V7);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V16);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V17);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V18);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V19);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V20);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V21);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V22);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V23);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V24);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V25);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V26);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V27);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V28);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V29);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V30);
+    tcg_regset_set_reg(tcg_target_call_clobber_regs, TCG_REG_V31);
+
     s->reserved_regs = 0;
     tcg_regset_set_reg(s->reserved_regs, TCG_TMP0);
     /* XXX many insns can't be used with R0, so we better avoid it for now */

tcg/s390x/tcg-target.h

@@ -37,11 +37,20 @@ typedef enum TCGReg {
     TCG_REG_R8,  TCG_REG_R9,  TCG_REG_R10, TCG_REG_R11,
     TCG_REG_R12, TCG_REG_R13, TCG_REG_R14, TCG_REG_R15,
 
+    TCG_REG_V0 = 32, TCG_REG_V1,  TCG_REG_V2,  TCG_REG_V3,
+    TCG_REG_V4,  TCG_REG_V5,  TCG_REG_V6,  TCG_REG_V7,
+    TCG_REG_V8,  TCG_REG_V9,  TCG_REG_V10, TCG_REG_V11,
+    TCG_REG_V12, TCG_REG_V13, TCG_REG_V14, TCG_REG_V15,
+    TCG_REG_V16, TCG_REG_V17, TCG_REG_V18, TCG_REG_V19,
+    TCG_REG_V20, TCG_REG_V21, TCG_REG_V22, TCG_REG_V23,
+    TCG_REG_V24, TCG_REG_V25, TCG_REG_V26, TCG_REG_V27,
+    TCG_REG_V28, TCG_REG_V29, TCG_REG_V30, TCG_REG_V31,
+
     TCG_AREG0 = TCG_REG_R10,
     TCG_REG_CALL_STACK = TCG_REG_R15
 } TCGReg;
 
-#define TCG_TARGET_NB_REGS 16
+#define TCG_TARGET_NB_REGS 64
 
 /* A list of relevant facilities used by this translator.  Some of these
    are required for proper operation, and these are checked at startup.  */
@@ -54,8 +63,9 @@ typedef enum TCGReg {
 #define FACILITY_FAST_BCR_SER         FACILITY_LOAD_ON_COND
 #define FACILITY_DISTINCT_OPS         FACILITY_LOAD_ON_COND
 #define FACILITY_LOAD_ON_COND2        53
+#define FACILITY_VECTOR               129
 
-extern uint64_t s390_facilities[1];
+extern uint64_t s390_facilities[3];
 
 #define HAVE_FACILITY(X) \
     ((s390_facilities[FACILITY_##X / 64] >> (63 - FACILITY_##X % 64)) & 1)
@@ -128,6 +138,27 @@ extern uint64_t s390_facilities[1];
 #define TCG_TARGET_HAS_muluh_i64      0
 #define TCG_TARGET_HAS_mulsh_i64      0
 
+#define TCG_TARGET_HAS_v64            HAVE_FACILITY(VECTOR)
+#define TCG_TARGET_HAS_v128           HAVE_FACILITY(VECTOR)
+#define TCG_TARGET_HAS_v256           0
+
+#define TCG_TARGET_HAS_andc_vec       0
+#define TCG_TARGET_HAS_orc_vec        0
+#define TCG_TARGET_HAS_not_vec        0
+#define TCG_TARGET_HAS_neg_vec        0
+#define TCG_TARGET_HAS_abs_vec        0
+#define TCG_TARGET_HAS_roti_vec       0
+#define TCG_TARGET_HAS_rots_vec       0
+#define TCG_TARGET_HAS_rotv_vec       0
+#define TCG_TARGET_HAS_shi_vec        0
+#define TCG_TARGET_HAS_shs_vec        0
+#define TCG_TARGET_HAS_shv_vec        0
+#define TCG_TARGET_HAS_mul_vec        0
+#define TCG_TARGET_HAS_sat_vec        0
+#define TCG_TARGET_HAS_minmax_vec     0
+#define TCG_TARGET_HAS_bitsel_vec     0
+#define TCG_TARGET_HAS_cmpsel_vec     0
+
 /* used for function call generation */
 #define TCG_TARGET_STACK_ALIGN		8
 #define TCG_TARGET_CALL_STACK_OFFSET	160

tcg/s390x/tcg-target.opc.h

@@ -0,0 +1,12 @@
+/*
+ * Copyright (c) 2021 Linaro
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or
+ * (at your option) any later version.
+ *
+ * See the COPYING file in the top-level directory for details.
+ *
+ * Target-specific opcodes for host vector expansion.  These will be
+ * emitted by tcg_expand_vec_op.  For those familiar with GCC internals,
+ * consider these to be UNSPEC with names.
+ */